Power Rankings

A sports rating system is a system that analyzes the results of sports competitions to provide ratings for each team or player. Common systems include polls of expert voters, crowdsourcing non-expert voters, betting markets, and computer systems. Ratings, or power ratings, are numerical representations of competitive strength, often directly comparable so that the game outcome between any two teams can be predicted. Rankings, or power rankings, can be directly provided (e.g., by asking people to rank teams), or can be derived by sorting each team's ratings and assigning an ordinal rank to each team, so that the highest rated team earns the #1 rank. Rating systems provide an alternative to traditional sports standings which are based on win–loss–tie ratios.

In the United States, the biggest use of sports ratings systems is to rate NCAA college football teams in Division I FBS, choosing teams to play in the College Football Playoff. Sports ratings systems are also used to help determine the field for the NCAA men's and women's basketball tournaments, men's professional golf tournaments, professional tennis tournaments, and NASCAR. They are often mentioned in discussions about the teams that could or should receive invitations to participate in certain contests, despite not earning the most direct entrance path (such as a league championship).

Computer rating systems can tend toward objectivity, without specific player, team, regional, or style bias. Ken Massey writes that an advantage of computer rating systems is that they can "objectively track all" 351 college basketball teams, while human polls "have limited value". Computer ratings are verifiable and repeatable, and are comprehensive, requiring assessment of all selected criteria. By comparison, rating systems relying on human polls include inherent human subjectivity; this may or may not be an attractive property depending on system needs.

History

Sports ratings systems have been around for almost 80 years, when ratings were calculated on paper rather than by computer, as most are today. Some older computer systems still in use today include: Jeff Sagarin's systems, the ''New York Times'' system, and the Dunkel Index, which dates back to 1929. Before the advent of the college football playoff, the Bowl Championship Series championship game participants were determined by a combination of expert polls and computer systems.

Theory

Sports ratings systems use a variety of methods for rating teams, but the most prevalent method is called a power rating. The power rating of a team is a calculation of the team's strength relative to other teams in the same league or division. The basic idea is to maximize the amount of transitive relations in a given data set due to game outcomes. For example, if A defeats B and B defeats C, then one can safely say that A>B>C.

There are obvious problems with basing a system solely on wins and losses. For example, if C defeats A, then an intransitive relation is established (A > B > C > A) and a ranking violation will occur if this is the only data available. Scenarios such as this happen fairly regularly in sports—for example, in the 2005 NCAA Division I-A football season, Penn State beat Ohio State, Ohio State beat Michigan, and Michigan beat Penn State. To address these logical breakdowns, rating systems usually consider other criteria such as the game's score and where the match was held (for example, to assess a home field advantage). In most cases though, each team plays a sufficient amount of other games during a given season, which lessens the overall effect of such violations.

From an academic perspective, the use of linear algebra and statistics are popular among many of the systems' authors to determine their ratings. Some academic work is published in forums like the MIT Sloan Sports Analytics Conference, others in traditional statistics, mathematics, psychology, and computer science journals.

If sufficient "inter-divisional" league play is not accomplished, teams in an isolated division may be artificially propped up or down in the overall ratings due to a lack of correlation to other teams in the overall league. This phenomenon is evident in systems that analyze historical college football seasons, such as when the top Ivy League teams of the 1970s, like Dartmouth, were calculated by some rating systems to be comparable with accomplished powerhouse teams of that era such as Nebraska, USC, and Ohio State. This conflicts with the subjective opinion that claims that while good in their own right, they were not nearly as good as those top programs. However, this may be considered a "pro" by non-BCS teams in Division I-A college football who point out that ratings systems have proven that their top teams belong in the same strata as the BCS teams. This is evidenced by the 2004 Utah team that went undefeated in the regular season and earned a BCS bowl bid due to the bump in their overall BCS ratings via the computer ratings component. They went on to play and defeat the Big East Conference champion Pittsburgh in the 2005 Fiesta Bowl by a score of 35-7. A related example occurred during the 2006 NCAA men's basketball tournament where George Mason were awarded an at-large tournament bid due to their regular season record and their RPI rating and rode that opportunity all the way to the Final Four.

Goals of some rating systems differ from one another. For example, systems may be crafted to provide a perfect retrodictive analysis of the games played to-date, while others are predictive and give more weight to future trends rather than past results. This results in the potential for misinterpretation of rating system results by people unfamiliar with these goals; for example, a rating system designed to give accurate point spread predictions for gamblers might be ill-suited for use in selecting teams most deserving to play in a championship game or tournament.

Rating considerations

Home advantage

When two teams of equal quality play, the team at home tends to win more often. The size of the effect changes based on the era of play, game type, season length, sport, even number of time zones crossed. But across all conditions, "simply playing at home increases the chances of winning." A win away from home is therefore seen more favorably than a win at home, because it was more challenging. Home advantage (which, for sports played on a pitch, is almost always called "home field advantage") is also based on the qualities of the individual stadium and crowd; the advantage in the NFL can be more than a 4-point difference from the stadium with the least advantage to the stadium with the most.

Strength of schedule

Strength of schedule refers to the quality of a team's opponents. A win against an inferior opponent is usually seen less favorably than a win against a superior opponent. Often teams in the same league, who are compared against each other for championship or playoff consideration, have not played the same opponents. Therefore, judging their relative win–loss records is complicated.

The college football playoff committee uses a limited strength-of-schedule algorithm that only considers opponents' records and opponents' opponents' records (much like RPI).

Points versus wins

A key dichotomy among sports rating systems lies in the representation of game outcomes. Some systems store final scores as ternary discrete events: wins, draws, and losses. Other systems record the exact final game score, then judge teams based on margin of victory